Mindcase for Real-Time Brand Monitoring

A product defect, a bad customer interaction, a screenshot that gets quote-tweeted — none of these start as a crisis. They start as one post, and whether they stay that way depends on how fast you notice. Checking your usual dashboard once a day misses the window where a single complaint is still just a single complaint.

If you're an agent (or building one) reading this rather than a human, the full machine-readable schema for every endpoint below lives at mindcase.co/skills.md.

What makes real-time monitoring different from routine listening?

It's not a different set of platforms — it's a different time window and a different starting point. Routine listening sweeps a keyword across a week and ranks what floats to the top. Real-time monitoring narrows that same sweep to the last hour and treats "nothing new" as the expected result, not a wasted check. It also splits into two distinct jobs: discovering posts you don't know about yet (Reddit, Twitter/X, Google News all support keyword search), and watching reaction to a post you already know about — your own company's Facebook post, for instance, which only supports a direct post URL, not a keyword search.

How do you pull the first signals across Reddit, Twitter/X, and Google News?

All three support a keyword and a short lookback window, so the same search runs across all three in parallel:

curl -X POST "https://api.mindcase.co/v1/data/reddit/posts/run?wait=true" \
-H "Authorization: Bearer $MINDCASE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "params": {
    "keyword": "acme water bottle leak",
    "sortBy": "new",
    "timeRange": "hour",
    "maxResults": 50
  }
}'

# $0.005 per post returned

Reddit Posts takes timeRange: "hour" directly. Twitter/X Tweets has no built-in time filter, but its query param accepts real search operators — since:2026-10-31_14:00:00_UTC narrows a search-operator query to whatever window you need. Google News goes further: timeframe: "1h" is a real supported option, built for exactly this — catching coverage before it's a day old.

How do you monitor reaction on your own Facebook post?

This is the one endpoint in this workflow that doesn't discover anything — it reads comments on a post URL you already have, which for brand monitoring usually means your own company page's most recent post:

curl -X POST "https://api.mindcase.co/v1/data/facebook/comments/run?wait=true" \
-H "Authorization: Bearer $MINDCASE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "params": {
    "post_url": "https://www.facebook.com/AcmeCo/posts/1234567890",
    "viewOption": "RECENT_ACTIVITY",
    "maxResults": 100
  }
}'

# $0.0025 per comment returned

viewOption: "RECENT_ACTIVITY" sorts by newest first instead of Facebook's default ranked order — the right choice when you're checking what just got posted, not what's most liked overall.

What can't you get from Mindcase?

Two cases.

Discovering an unknown Facebook post. Facebook Comments takes a post URL, not a keyword — there's no way to search Facebook itself for "posts mentioning Acme" the way Reddit, Twitter/X, and Google News all support. If the mention isn't on a post you already know about, this endpoint can't find it.

A severity or sentiment score. None of these four endpoints label a post as good, bad, or urgent — you get the raw text, timestamp, and platform-native engagement numbers (upvotes, views, comment counts). Deciding that a spike of new posts in the last hour is worth escalating is a judgment call you make after retrieval, typically by handing the batch to an LLM, not something baked into any of these responses.

Which endpoint should you use for which job?

A single hourly check — 50 Reddit posts, 100 tweets, 20 news articles, and 100 comments on your latest Facebook post — costs $0.25 + $0.025 + $0.04 + $0.25, or $0.565 per pass. Running that check every hour during a live incident for eight hours costs under $5 total, which is the kind of number that makes "just check again in ten minutes" a reasonable thing to automate rather than something a person has to remember to do.

How do you use this data responsibly?

Speed cuts both ways. A handful of posts in the first hour is not automatically a trend — verify a claim before acting on it publicly, and don't mistake an hour of elevated volume for a settled fact about scale. Everything here is public post data with no login required, but treat identifying details in individual posts the same way you would in routine listening: don't publish someone's username or profile alongside a complaint you're escalating internally, and check your own legal or comms team's guidance before responding publicly to anything pulled this way.

FAQ